Table 2.

Optimal results of Models 1–3 on the validation set

Model name	Construction method	Optimal algorithm	Number of optimal feature	Negative training n sample	Positive training n sample	Negative validation sample	Positive validation sample	Validation accuracy	ValidationF1 score
Model1	Proportional division	GBDT	9	175,496	10,756	43,873	2690	0.96	0.78
Model2	Upsampling	RF	3	175,395	172,209	43,974	42,927	0.99	0.99
Model3	Downsampling	GBDT	24	10,784	10,729	2662	2717	0.84	0.84